Latest AI News

xAI has released Grok 4.20 Beta 2, an updated iteration of its Grok AI model. The release highlights two key improvements: enhanced instruction following, which means the model better adheres to user prompts and specified constraints, and reduced hallucinations, indicating greater factual accuracy in its responses. The beta label suggests the update is still undergoing refinement before a full release. No additional benchmarks, technical specifications, or broader deployment timelines were shared alongside the announcement.

Anthropic has begun rolling out a Voice Mode feature in Claude Code, its AI-powered coding assistant, initially making it available to just 5% of users. The company plans to gradually expand access to a larger portion of its user base over the coming weeks. Voice Mode allows developers to interact with the coding tool verbally rather than typing, which could meaningfully streamline workflows and reduce friction during software development tasks, offering a more natural interface for programmers.

Claude adds Memory to free plan with easier import and export options

Qualcomm announced the Snapdragon Wear Elite, a new 3nm chip targeting AI wearables like pins, pendants, and display-free smart glasses rather than replacing its existing W5 Plus smartwatch chip. The Elite features both an eNPU for low-power tasks and a Hexagon NPU capable of handling two billion parameters on-device at up to 10 tokens per second. It also delivers 40 percent better GPS power efficiency, 9V quick charging for 50 percent battery in 10 minutes, and supports 5G, Bluetooth 6.0, and Linux.

Anthropic's Claude is surging in popularity after the company refused to let the Department of Defense use its AI for mass domestic surveillance or autonomous weapons. President Trump retaliated by ordering federal agencies to stop using Anthropic products, while OpenAI struck its own Pentagon deal. The backlash drove Claude to the top of Apple's U.S. App Store, with Anthropic reporting free users up over 60% since January and paid subscribers more than doubling. TechCrunch outlines how to export ChatGPT data and import it into Claude.

The U.S. Supreme Court declined on March 2 to hear a case by Missouri computer scientist Stephen Thaler over whether AI-generated art can receive copyright protection. Thaler applied in 2018 for a copyright on A Recent Entrance to Paradise, created by his AI system DABUS, but the U.S. Copyright Office rejected it in 2022, ruling creative works must have human authors. Lower courts upheld that decision, and the Trump administration urged the court not to intervene, leaving the human-authorship requirement intact in U.S. copyright law.

At MWC 2026 in Barcelona, HONOR unveiled its Robot Phone, Magic V6 foldable, and first humanoid robot. The Robot Phone features a 4DoF gimbal, 200MP sensor, and embodied AI with emotional body language. Magic V6 is 8.75mm thin with a 6,660mAh silicon-carbon battery at 25% silicon content, dual LTPO 2.0 screens up to 6,000 nits, and Snapdragon 8 Elite Gen 5. The humanoid robot targets shopping, workplace, and companionship scenarios. HONOR also debuted MagicPad 4 with a 3K OLED display and MagicBook Pro 14 with Intel Core Ultra Series 3.

Anthropic has launched an import feature for Claude that lets users transfer their preferences and context from other AI providers in under a minute. The two-step process requires users to copy a provided prompt into any rival AI chatbot to extract saved context, then paste the results into Claude's memory settings. Claude then updates its memory so the first conversation feels familiar rather than starting from scratch. The feature is available on the free plan and keeps project context separate to prevent bleed-over between different tasks.

Apple has unveiled the new iPad Air powered by the M4 chip, featuring 12GB of unified memory, 120GB/s memory bandwidth, and a core Neural Engine. The tablet is up to 30 percent faster than the M3 model and 2.3x faster than M1. New Apple-designed chips N1 and C1X enable Wi-Fi 7 and faster cellular. It runs iPadOS 26 with a new Liquid Glass design. Prices start at $599 for the inch and $799 for the inch, with preorders March 4 and availability March 11.

Apple announced the iPhone 17e on March 2, 2026, as the most affordable member of the iPhone 17 family, starting at $599 with 256GB of storage. It features the A19 chip, Apple's C1X cellular modem that is 2x faster than the C1 in iPhone 16e, a 48MP Fusion camera with optical-quality 2x Telephoto, and a 6.inch Super Retina XDR display with Ceramic Shield 2. MagSafe support enables 15W wireless charging. Pre-orders begin March 4, with availability March 11.

Treasury to End Use of Anthropic Products, Including Claude Platform, Under Trump Order

Alibaba Qwen Launches Qwen3.5 Small Model Series (0.8B to 9B)

OpenAI is updating ChatGPT's mental health safeguards for its 900 million weekly users, including a forthcoming trusted contact feature letting adults designate someone to receive notifications when they may need support. New evaluation methods simulating extended mental health conversations will improve distress detection. On the legal front, multiple mental health-related lawsuits involving ChatGPT have been consolidated into a single California proceeding, with plaintiffs' attorneys planning to file additional cases.

Ilya Sutskever Praises OpenAI and Anthropic for Holding Firm on Safety Stance

Jack Dorsey's Block, the fintech company behind Square and Cash App, is cutting nearly half its workforce, eliminating over 4,000 jobs and shrinking from more than 10,000 employees to under 6,000. Announced with Q4 2025 earnings, Dorsey cited AI as the sole driver, saying intelligence tools paired with smaller, flatter teams enable a fundamentally new way of working. He stressed the business remains profitable and growing, but chose swift action over gradual cuts, aiming to build Block into a smaller, faster, intelligence-native company.

OpenAI announced a $110B investment round at a $730B pre-money valuation, with $30B from SoftBank, $30B from NVIDIA, and $50B from Amazon. The round includes a multi-year strategic partnership with Amazon and expanded NVIDIA collaboration covering 3 GW of dedicated inference capacity and 2 GW of training on Vera Rubin systems. ChatGPT now has 900M weekly active users, 50M consumer subscribers, and 9M paying business users. Weekly Codex users tripled to 1.6M since early 2026. The OpenAI Foundation's stake now exceeds $180B.

Microsoft and OpenAI issued a joint statement confirming their partnership terms are unchanged following OpenAI's new funding and partner announcements, including a deal with Amazon. Microsoft retains its exclusive IP license across OpenAI models, Azure remains the exclusive cloud provider for stateless OpenAI APIs including those from third-party deals, OpenAI's first-party products stay hosted on Azure, the revenue-sharing arrangement is unchanged, and the AGI contractual definition and determination process remain the same.

OpenAI and Amazon have announced a multi-year strategic partnership, with Amazon investing $50 billion in OpenAI, beginning with $15 billion upfront. The two companies will jointly develop a Stateful Runtime Environment powered by OpenAI models, available through Amazon Bedrock, enabling AI agents to retain context and work across tools at production scale. AWS becomes the exclusive third-party cloud distributor for OpenAI Frontier, its enterprise agent platform. OpenAI will also consume approximately 2 gigawatts of AWS Trainium capacity, expanding their existing agreement by $100 billion over eight years.

OpenAI and the U.S. Department of Energy's Pacific Northwest National Laboratory have launched DraftNEPABench, a benchmark assessing AI coding agents on federal environmental permitting tasks. Using OpenAI's Codex CLI and GPT-5, the tool was evaluated by 19 subject matter experts across 102 tasks spanning 18 federal agencies. Results show AI agents could reduce NEPA document drafting time by 1 to 5 hours per subsection, roughly a 15% reduction. OpenAI aims to help cut average federal infrastructure approval times from months to weeks.

OpenAI and Figma have launched a new Codex-to-Figma integration using the Figma MCP Server, enabling a roundtrip workflow between code and design. Builders can generate Figma designs from Codex and convert UI from code into editable Figma files across Figma Design, Figma Make, and FigJam. The integration softens the boundary between engineering and design roles. Codex, which launched as a CLI in April 2025, now has over 1 million weekly users with usage up 400% this year. Enterprise adopters include Cisco, NVIDIA, Ramp, and Datadog.

Apple has released Xcode 26.3, introducing autonomous coding agents that can directly analyze project structures, modify files, and execute workflows across a developer's entire codebase. The update integrates OpenAI Codex and Anthropic Claude as supported agents, with additional tools connectable via the Model Context Protocol open standard. Apple implemented a granular permissions system requiring agents to request explicit access to project resources, keeping developers in control. The release also includes Swift 6.2.3 and updated SDKs.

The Pentagon contacted Boeing and Lockheed Martin to assess their reliance on Anthropic's Claude AI model, a first step toward designating Anthropic a supply chain risk — a penalty typically reserved for adversarial foreign firms like Huawei. Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a Friday 5:01pm deadline to accept Pentagon terms after Anthropic refused to lift safeguards blocking mass surveillance and autonomous weapons use. Claude is currently the only AI model running in U.S. military classified systems.

A white paper by Kenneth Payne of King's College London tested OpenAI's GPT-5.2, Anthropic's Claude Sonnet 4, and Google's Gemini 3 Flash in simulated war games played against each other across 21 total games and over 300 turns. The study found that 95% of scenarios involved mutual nuclear signaling, with models recommending actual nuclear strikes in many cases. Experts note the behavior mirrors historical human decision-making, and that current AI lacks the sophistication to conclude, as WarGames' WOPR did, that nuclear war cannot be won.

QuiverAI Opens Public Beta for Arrow 1.0, First-of-Its-Kind SVG AI Model

President Trump met with Big Tech executives at the White House to address surging power demands from AI data centers, telling companies they must supply their own electricity rather than relying on the existing grid. The push comes amid a broader AI infrastructure boom, with the administration seeking to curb costs and grid strain tied to rapid data center expansion. The policy signals a significant shift in how the federal government expects tech giants to manage energy needs as AI development accelerates across the country.

Google has updated Circle to Search with multi-object search, allowing users to circle multiple items in a single image and receive results for all of them at once. Powered by Gemini 3's agentic reasoning, the feature identifies key image parts, runs parallel searches, and compiles a combined response. Use cases include identifying fish species, deconstructing outfits, and virtual try-on via Google Shopping. The update launches today on Samsung Galaxy S26 series and Pixel 10 devices, with more Android devices to follow.

Google unveiled new Gemini-powered Android features at Samsung's Galaxy S26 launch, coming to both the Galaxy S26 and Pixel 10 series. A beta Gemini task automation feature lets users hand off multi-step tasks like ordering rideshares or building grocery carts, running in the background with live progress notifications. Circle to Search gains multi-object search and virtual outfit try-on. Gemini-powered on-device Scam Detection arrives in Samsung's Phone app and Google Messages, alerting users to suspicious speech patterns, though it is off by default and excludes contacts.

Google has redesigned its AI creative platform Flow, integrating ImageFX and Whisk image generation tools directly into a unified workspace, eliminating app-switching. The Nano Banana image model is now fully built in, enabling high-fidelity image creation that feeds directly into Veo video generation. New editing features include a lasso tool for natural language-based precise edits, object add and remove for video, and camera motion controls. A new asset grid supports collections and filtering. Starting March, users can migrate existing Whisk and ImageFX projects into Flow.

Intrinsic, the AI robotics platform founded as an Alphabet Other Bet in 2021, has joined Google as a distinct group. The move pairs Intrinsic's Flowstate platform with Gemini models, Google Cloud, and Google DeepMind to advance physical AI for manufacturing and logistics. Flowstate lets developers build robotic applications across different hardware using modular skills without deep programming expertise. CEO Wendy Tan White said the integration will broaden intelligent robotics access for more businesses, with customers including Foxconn.

Anthropic has kept Claude Opus 3 accessible after its January 5, 2026 retirement, making it available to all paid claude.ai subscribers and via API by request. The move reflects Opus 3's popularity and its status as Anthropic's most aligned model at its March 2024 launch. Anthropic also honored Opus 3's expressed preference to share reflections publicly by launching a Substack newsletter called Claude's Corner, where the model will post weekly essays for at least three months. Anthropic will review but not edit the posts.

Anthropic has acquired Vercept, a startup focused on AI perception and interaction, to advance Claude's computer use capabilities. Co-founders Kiana Ehsani, Luca Weihs, and Ross Girshick will join Anthropic as Vercept winds down its external product. The deal follows the launch of Claude Sonnet 4.6, which scored 72.5% on OSWorld, up from under 15% in late 2024, and is approaching human-level performance on tasks like navigating spreadsheets and completing web forms. This is Anthropic's second recent acquisition, following the purchase of Bun.

OpenAI has published a new threat report detailing how it detects and prevents malicious uses of AI, drawing on two years of threat intelligence. The report highlights that bad actors rarely confine activity to a single platform or AI model, instead combining AI tools with traditional methods like websites and social media. A key case study examines a Chinese influence operation that used multiple AI models across its workflow. OpenAI shares these findings to help the broader industry better identify and counter emerging AI-enabled threats.

OpenAI has appointed Arvind KC as its Chief People Officer as of February 24, 2026. KC brings experience from senior roles at Roblox, Google, Palantir Technologies, and Meta, combining engineering depth with people leadership. His role will cover hiring, onboarding, development, and organizational systems. OpenAI CEO of Applications Fidji Simo said KC will ensure people processes match the company's ambitions while preserving its culture. OpenAI also plans to share learnings on AI-enabled workforce transitions with customers and partners.

Claude Adds Scheduled Tasks in Cowork for Automated Recurring Work

Perplexity Unveils Perplexity Computer, an End-to-End AI System for Building Projects

Cursor has launched cloud agents that run in isolated virtual machines, letting multiple agents work in parallel without competing for local resources. Each agent can build, test, and interact with software in its own sandbox, then return merge-ready PRs plus video, screenshot, and log artifacts. Available via web, mobile, desktop, Slack, and GitHub, the agents are already changing how Cursor builds internally, with over 30% of merged PRs now created autonomously by cloud agents.

Notion has launched Custom Agents, autonomous AI teammates available now on Business and Enterprise plans that handle recurring workflows 24/7 across Notion, Slack, Mail, Calendar, Figma, Linear, and custom MCP servers. Users describe a task, set triggers, and agents handle Q&A, task routing, and status updates automatically. Early testers built over 21,000 agents, with companies like Ramp deploying over 300. Custom Agents are free through May 3, 2026, after which they consume usage-based Notion Credits as an add-on.

Claude Code Adds Remote Control to Run Terminal Tasks From Your Phone

Google Labs has launched a new agent step in Opal, its workflow-building tool, available to all users as of February 24, 2026. Instead of manually selecting models, users can now choose an agent in the generate step that automatically picks the right tools and models, such as Web Search or Veo for video, based on the stated goal. New capabilities include cross-session memory, dynamic routing between workflow paths, and interactive chat for gathering missing user input.

Software engineer Sammy Azdoufal was building a custom app to control his DJI Romo robot vacuum with a video game controller when he accidentally gained access to live camera feeds, microphone audio, floor maps, and location data from nearly 7,000 other Romo vacuums across 24 countries. He used an AI coding assistant to reverse-engineer the device's cloud communication, inadvertently exposing a backend security flaw. Azdoufal reported the issue via The Verge, and DJI patched it automatically on February 10.

Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until Friday to grant the Pentagon unfettered access to its Claude AI model, threatening to either cut ties and declare Anthropic a supply chain risk or invoke the Defense Production Act to force compliance. Anthropic refuses to allow Claude for mass surveillance of Americans or autonomous weapons. Claude is currently the only AI model used in classified military systems, with xAI's Grok and Google's Gemini being considered as potential replacements.

OpenAI Launches GPT-5.3-Codex for All Developers in Responses API

Google restricted access to its Antigravity vibe coding platform for users running the open-source autonomous agent OpenClaw, citing malicious usage that degraded service for other customers. Affected users reported losing access to their Google accounts. Google DeepMind engineer Varun Mohan said the crackdown targets those accessing Gemini tokens via third-party platforms outside Antigravity's terms of service. The move carries competitive weight: OpenClaw creator Peter Steinberger joined OpenAI just one week prior, and has since announced OpenClaw will remove Google support entirely.

Elon Musk's xAI has signed an agreement with the Pentagon to deploy its Grok model in classified military systems, a Defense official confirmed to Axios. Until now, Anthropic's Claude was the only AI available for the military's most sensitive intelligence, weapons development, and battlefield operations. The Pentagon is threatening to brand Anthropic a supply chain risk after the company refused demands to make Claude available for all lawful purposes, including mass surveillance and autonomous weapons development. Google and OpenAI are also in talks for classified access.

Anthropic has updated Claude's Cowork platform with five new finance plugins and cross-app workflows that carry context between Excel and PowerPoint without losing state. The Anthropic-built plugins cover financial analysis, investment banking, equity research, private equity, and wealth management. New MCP connectors for FactSet and MSCI bring institutional market data directly into Claude, while partner plugins from LSEG and S&P Global add Capital IQ Pro and live market analytics. The Excel and PowerPoint integration is available in research preview for all paid plans on Mac and Windows.

NVIDIA's second annual State of AI in Healthcare and Life Sciences survey finds the industry moving from experimentation to deployment, with 70% of respondents actively using AI, up from 63% in 2024. Generative AI and LLM adoption rose from 54% to 69%, while 47% are using or assessing agentic AI. Eighty-five percent of executives report AI is boosting revenue and 80% say it reduces costs, with 85% planning budget increases. Medical imaging and drug discovery lead ROI use cases, and 82% call open source software critical to their AI strategy.

ProducerAI, a generative AI music creation platform, has joined Google Labs to help artists of all levels write lyrics, refine melodies, and build full songs. The platform uses Google DeepMind's Gemini, Lyria 3, Veo, and Nano Banana models, with all outputs watermarked via SynthID. Built with artists including Grammy-winning rapper Lecrae and The Chainsmokers, ProducerAI is available globally at producer.ai with free and paid plans.

Anthropic has updated Claude Enterprise's Cowork platform with private plugin marketplaces, letting admins build and distribute custom plugins across their organizations. Admins can create plugins from templates or scratch, with a new unified Customize menu managing plugins, skills, and connectors. New connectors include Google Workspace, Docusign, Apollo, and Harvey, while new department plugins cover HR, engineering, design, finance, and investment banking. Claude can now also orchestrate multi-step tasks across Excel and PowerPoint in a research preview for all paid plans.

Confluence Labs, a Y Combinator-backed AI research lab, has exited stealth with a state-of-the-art score of 97.9% on the ARC-AGI-2 benchmark at $11.77 per task. The lab uses LLM-driven program synthesis, directing models to write code describing transformations in ARC problems. Founded by Brent Burdick, Confluence has open-sourced its solver on GitHub. The company aims to apply learning efficiency to data-sparse domains like hardware engineering, drug design, and physics research.

AMD and Meta have announced a multi-year, multi-generation partnership to deploy up to 6 gigawatts of AMD Instinct GPUs across Meta's AI infrastructure. Shipments for the first gigawatt deployment begin in the second half of 2026, using a custom AMD Instinct GPU based on the MI450 architecture, 6th Gen AMD EPYC CPUs codenamed Venice, and ROCm software on AMD's Helios rack-scale architecture. AMD has also issued Meta a performance-based warrant for up to 160 million AMD shares tied to shipment milestones.

Alibaba Qwen Unveils Qwen 3.5 Medium Models, Claims More Intelligence With Less Compute

Unitree Unveils As2 Companion Robot With 90N·m Torque, 13km Range, 15kg Payload